Released in April, 2012, Geosoft’s VOXI Earth Modelling service gives exploration geophysicists the ability to convert magnetic and gravity data directly into useful 3D Earth models that can be integrated with other 3D exploration data.
Creating the cloud-based VOXI inversion service took us over three years, and required the work of 29 people on a number of development teams. This is the story of how we did it.
3D Earth Modelling and Exploration
Three-dimensional Earth modelling of geophysical data has been an important part of exploration geophysics for over a decade, allowing explorers to transform 2D data into detailed, subsurface 3D models useful for defining exploration targets in areas of complex geology.
The science behind 3D inversions has been in development within academic environments even longer. The University of British Columbia Geophysical Inversion Facility (UBC-GIF), a leader in this area of research, was founded in 1989. The UBC-GIF consortium for Joint and Cooperative Inversion of Geophysical and Geological Data (JACI) pioneered the developed practical inversion algorithms that have been used in real exploration since the 1990s.
It’s our view, however, that the full potential for 3D inversion in exploration has continued to be limited both by the complexity of the tools and the need for powerful computers. With VOXI, our goal was to address both of these challenges and make 3D Earth Modelling accessible to more explorers and be more impactful to exploration success.
Behind VOXI’s development
VOXI development began in 2009, the year Dr. Robert G. Ellis joined Geosoft as a Senior Scientist for Earth Modelling. As a co-founder of the UBC-GIF, and a collaborator with UBC and other modelling groups over 20 years, Rob is a renowned expert in 3D geophysical inversion. Rob and Geosoft embarked on a shared mission: to build the next generation 3D earth modelling system that would combine Rob’s accumulated knowledge with the latest available algorithms and commercial development techniques in use at Geosoft.
We also saw the opportunity to harness cloud computing to provide super-computing power to everyone. The cloud presented us with both challenge and opportunity. Using cloud computing meant interpreters would not be limited by the processing power of their computers, which is a huge benefit for process-intensive applications like 3D inversion. Our challenge was to scale VOXI so that hundreds of geophysicists could run large inversions at the same time, and from anywhere in the world.
Scaling for cluster computing
With Rob's knowledge of the science of 3D inversion, in less than 6 months we were able to build the latest generation of inversion from both open-source libraries and our own original development. Then came what turned out to be the really hard work – making the system work on multiple-cores in a cloud environment, and making it easy to use.
Our first scaling work was done on a conventional 96-core computing platform hosted in the “cloud” by Rackspace in the USA. All our multi-threading and cluster-computing development was proven here, and we were able to achieve some impressive performance numbers with Ivanhoe Mines as a key early collaborator accessing the service from Australia. But we still faced the challenge of how to make VOXI scale so that many geophysicists could run large inversions at the same time, and from anywhere in the world. For that we turned to the Microsoft Azure platform.
True Cloud computing
Azure was important to us in a number of ways. First and foremost was the level of commitment that Microsoft has made to Azure, which has the ability to support our exploration customers anywhere on earth with 6 international data centers, each with up to 3 million processing cores. Microsoft also had both the business models and the tools we needed to be able to scale as demand required.
But developing for a true cloud platform is not straight-forward, as we would find out. Building a multi-core cloud-based algorithm is very different from building a program that runs on a workstation or local cluster. The effort to re-engineer VOXI for Azure took us most of a year to get right.
Making VOXI usable 
And as any commercial software company knows, building software to be usable is a challenge in its own right. We worked very hard with our collaborators over two years to design and improve the interface so that everything would work as smoothly as possible and fit within an explorer's natural workflow. The user-interface development work required yet another full-time development team.
Summary of the benefits
We feel the outcome has been well-worth the journey, and we are excited about the benefits we see for exploration:
- All explorers have super-computer access to build large models in minutes.
One early VOXI customer compared the VOXI service performance to his current desktop modelling program. Given exactly the same model, what took four hours with his current software was completed by VOXI in under 2 minutes - more than 100 times faster.
- 3D Earth modelling is now practical and user-friendly for all geophysicists.
VOXI runs with a minimum of user effort directly from Geosoft Oasis montaj, an industry-standard platform for working with geophysical data. It just works, naturally, and seamlessly.
What’s Next?
VOXI has been our largest and most comprehensive development to date. The work we have done with VOXI lays the foundation we need to build more useful cloud capabilities for explorers in the years to come, including more 3D modelling methods, support for larger models and new ways to harness cloud computing to benefit exploration.